Add commentEdit

Website stack

How the TCF, campaign and Not Bastards websites are built and deployed. Part of Infrastructure. What each site is for, and its status, is on Websites. Hosting providers are on Hosting.

Stacks

There are two stacks:

Not every site is built by GitLab CI: graphic.notbastards.com is deployed by hand with wrangler pages deploy public, and jonkerbijl.racing is built by Cloudflare Pages Git integration (build command npm test, output directory public).

Sites

State on 10 October 2026. Most repositories are in the tcfbv/brand group and checked out under umbrella/; the exceptions are named in the table.

Site Stack Fonts Analytics @tcfbv/web-tracking Tests in CI
conveniencefactory.com Vue app, prerendered; English at /, Dutch at /nl/ Self-hosted Umami Yes Yes
ctotehuur.nl Vue app, prerendered Self-hosted Umami Yes Yes
shitwemoetensoftwarebouwen.nl Vue app, prerendered Self-hosted (Inter via Fontsource) Umami Yes Yes
shitwemoetenietsmetai.nl Vue app, prerendered Self-hosted (Inter via Fontsource) Umami Yes Yes
productbuilder.dev Static HTML Self-hosted Umami Yes Yes
prototypeinaday.dev Static HTML Self-hosted Umami No Yes
mvpinaweek.dev Static HTML Self-hosted Umami No Yes
brand.conveniencefactory.com Static, built by scripts/build.mjs No No
notbastards.com Vue app, prerendered; build also generates the manifesto PDF, only when its source changes Umami No Yes
intelcountdown.com Vue app, prerendered (no DaisyUI), with Pages Functions; repo tcfbv/product/intelcountdown.com Umami No Yes
ip.notb.st Vue app, prerendered, with Pages Functions; repo tcfbv/ip-website, checked out at ~/git/ip-website Self-hosted (Fontsource) Umami No Yes
graphic.notbastards.com Static HTML in website/public/ of tcfbv/product/idrawgraphic-rescue; deployed by hand Self-hosted Umami No npm test before the manual deploy
jonkerbijl.racing Static HTML in public/; repo mdbijl/jonkerbijl.racing; Cloudflare Pages Git integration Self-hosted (Orbitron, Saira) No Yes (Pages build command npm test)

Umami is the cookieless, self-hosted analytics on the Hetzner VPS; the script loads from a.notb.st. @tcfbv/web-tracking (repository tcfbv/infra/web-tracking) adds the channel code to mail links and sends the contact-mail and contact-phone events; see Shit* measuring.

Findability files

Sites with “Yes” under findability files have:

On conveniencefactory.com the language toggle links / and /nl/ and stores the choice under tcf-locale-choice in localStorage. On /, a script in the page head sends a stored Dutch choice to /nl/; with no stored choice, a Dutch browser also goes to /nl/. Crawlers do not run the script.

Cloudflare bot settings

Per zone, under Security, Settings, filtered on Bot traffic:

AI Crawl Control shows which AI crawlers visit each site and which status they get.

Redirects between hosts are Cloudflare Redirect Rules (Rules, Redirect Rules) on the zone of the host that redirects. Each is a dynamic redirect to concat("https://<target>", http.request.uri.path) with status 301 and “Preserve query string” on, so path and query are kept and http goes straight to https. Checked on 9 October 2026:

Host Redirects to
www.conveniencefactory.com conveniencefactory.com
www.shitwemoetensoftwarebouwen.nl shitwemoetensoftwarebouwen.nl
www.shitwemoetenietsmetai.nl shitwemoetenietsmetai.nl
www.productbuilder.dev productbuilder.dev
www.prototypeinaday.dev prototypeinaday.dev
www.mvpinaweek.dev mvpinaweek.dev
www.notbastards.com notbastards.com
notb.st (apex only; ip.notb.st and a.notb.st are not redirected) notbastards.com
www.jonkerbijl.racing jonkerbijl.racing
www.intelcountdown.com intelcountdown.com (by the site’s own functions/_middleware.js, not a Redirect Rule)
shitwemoetensoftwaregaanbouwen.nl and its www shitwemoetensoftwarebouwen.nl
prototypeineendag.nl prototypeinaday.dev
mvpineenweek.nl mvpinaweek.dev
medtech.build conveniencefactory.com (HTTP 302; parked for now)

The target must start with https://. Without it, Cloudflare treats the target as a path on the same host and the redirect loops. _redirects in a Pages project only handles paths, not hosts.

New site checklist

Follow these steps to make a new site findable in search engines and AI tools. They repeat what was done for the seven sites in October 2026. The sections above give the details.

  1. HTML that crawlers can read. Static HTML is fine as it is. Prerender a Vue app: vite build, then an SSR build, then scripts/prerender.mjs with renderToString, which writes one index.html per route. Check with curl https://<domain>/: the response must contain the page text, not only <div id="app">.
  2. Files in public/:
    • robots.txt: all crawlers allowed, AI crawlers named explicitly, and a Sitemap: line.
    • sitemap.xml: every real URL, with <lastmod>.
    • 404.html: remove the SPA fallback /* /index.html 200 from _redirects.
    • llms.txt: a short summary with links, in the llmstxt.org format.
  3. Head tags on every page:
    • its own <title>, and a meta description of at most 155 characters (with “excl. VAT” next to any price);
    • a canonical link to the bare domain with a trailing slash, and og:title, og:description and an og:url that matches the canonical;
    • the correct <html lang>. A bilingual site gets separate URLs (/ and /nl/) with hreflang.
  4. JSON-LD:
    • a WebSite and a Service, whose provider/publisher is the TCF Organization with its type, name, url and "@id": "https://conveniencefactory.com/#organization";
    • an Offer only where the page shows a price;
    • an FAQPage only where the page shows the same questions and answers.
  5. Question pages:
    • Pick 2–3 buying-intent questions: cost, speed, what you get.
    • Give each question its own page at /<slug>/ with a breadcrumb, the question as H1, and a 40–60 word answer with a number directly below it.
    • Add a contact CTA and a “more questions” list, plus WebPage, BreadcrumbList and FAQPage (one question) in the JSON-LD.
    • The home page only links to the question pages, from the footer.
    • Add every question page to sitemap.xml and llms.txt.
  6. PDFs (optional): commit the PDF in public/documents/, add *.pdf binary to .gitattributes, and let scripts/copy-documents.mjs refresh it from tcfbv/sales/services when the source is newer. Link it only from the cost answer.
  7. Tests and CI: npm test checks the built dist/ (files, head tags, JSON-LD, links), and CI runs it before deploy. Push to development to test on dev.<domain>, then merge to main.
  8. Cloudflare, per zone:
    • AI bot policies: allow Search, Agent and Training.
    • Turn off Bot Preference Sync, Bot Fight Mode and AI Labyrinth.
    • Add a 301 Redirect Rule for www and any alias domain (see Cloudflare bot settings).
  9. Search engines:
  10. Check after go-live:
    • curl /robots.txt, /sitemap.xml, /llms.txt, an unknown path (expect 404) and the www host (expect 301).
    • After a few days, check AI Crawl Control and Search Console.

Repo layout

The umbrella repo (tcfbv/brand/umbrella) pulls each site in as a git submodule. The full proposed GitLab group layout (public, brand, product, client, project, infra) is documented in umbrella/REPO_STRUCTURE.md.

Deployment